The purpose of this study was to gain a better understanding of speech development across successive generations of heritage language users, examining how cross-linguistic, developmental and socio-cultural factors affect stop consonant production.

To this end, we recorded Sylheti and English stop productions of two sets of Bangladeshi heritage families: (1) first-generation adult migrants from Bangladesh and their (second-generation) UK-born children, and (2) second-generation UK-born adult heritage language users and their (third-generation) UK-born children.

The data were analysed auditorily, using whole-word transcription, and acoustically, examining voice onset time. Comparisons were then made in both languages across the four groups of participants, and cross-linguistically.

The results revealed non-native productions of English stops by the first-generation migrants but largely target-like patterns by the remaining sets of participants. The Sylheti stops exhibited incremental changes across successive generations of speakers, with the third-generation children’s productions showing the greatest influence from English.

This is one of few studies to examine both the host and heritage language in an ethnic minority setting, and the first to demonstrate substantial differences in heritage language accent between age-matched second- and third-generation children. The study shows that current theories of bilingual speech learning do not go far enough in explaining how speech develops in heritage language settings.

These findings have important implications for the maintenance, transmission and long-term survival of heritage languages, and show that investigations need to go beyond second-generation speakers, in particular in communities that do not see a steady influx of new migrants.